New overlay filter.  Thanks to astro for the idea and Mats for the direction :)

Still needs a little polish and maybe some thread safety.

The overlay filters looks like it will be loads of fun

# import desired filters
from org.myrobotlab.opencv import OpenCVFilterFaceDetect
from org.myrobotlab.opencv import OpenCVFilterOverlay

# start an opencv service
opencv = Runtime.start("opencv","OpenCV")

# add face detect
opencv.addFilter(OpenCVFilterFaceDetect("facedetect"))

overlay = OpenCVFilterOverlay("overlay")
# add first overlay at 0.3 alpha (text on transparent background)
overlay.addImage("overlay1.png", 0.3)
# add second overlay at 0.5 alpha (red square)
overlay.addImage("red.png", 0.5)

I think it would be fun to add a method which would accept a variable, which would be printed out - and if the variable changed, the text would change.  This might be easier in some cases rather than requiring the swapping of image overlays.

 

astro

7 years 1 month ago

Amazing!!

Last night I installed eclipse, in the weekend I will try to learn how to use it.

:)

astro.-

Gareth

7 years 1 month ago

Hasta la vista, baby.

GroG

7 years 1 month ago

In reply to by astro

Very cool idea Astro,
Unfortunately, it appears to be treating your 4 channel transparency png as a 3 channel image with no transparency.  Which makes all white lines disappear.

"Loading" a 4 channel png is not a problem - but the webcam image comes in as 3 channel and the "merge" blows up.  I have some leads of how to fix that but its "non" trivial.

The above is 
 

   filter.addImage("red.png", 0.4);
   filter.addImage("overlay_640x480.png", 0.2);
   
The white has completely disappeared, because "white" is chosen to replace the transparency :(

You either have to create graphics with non-white graphics, or wait till I figure out the 4/3 channel stuff.
 
As for text, I'm more interested in using OpenCv text facilities rather than, writing fonts to images and overlaying images.
 
The API I'm think about creating is one like this :

filter.addText("{key}", x, y, size, "{initial text}")

filter.addText("scanmode", 600, 30, 20, "SCAN MODE NODE)
filter.addText("assessment", 600, 40, 20, "ASSESSMENT NONE")

filter.addText("fit", 600, 50, 20, "FIT PROBABILITY 0.0")
 
Then through events and callbacks this data could be updated
 
filter.updateText("{key}",{newText})
filter.updateText("scanmode","SCAN MODE 34982930")
filter.updateText("assessment","ASSESSMENT THREAT")
filter.updateText("fit","FIT PROBABILITY 0.98")
 
make sense ?
 

If you want to send me another one with "black" vs white - we could see how it looks

Oh, okay. I get it.
That's how I thought it would look.

I can do a TIF with alpha channel, but I do not know if opencv can handle it. I am really looking for it to be white so that it is as similar as possible to the original.

Here in black and fonts:
http://www.polibol.com.ar/astro/

Real text and data with Opencv would be great!!!

I was thinking about an alternative way and last night I started playing with Rainmeter. It opens a lot more possibilities to the imagination and with animated vectors , only it does not have MRL data, I read that it is possible to make plugins that read variables but I do not know how to program in that.

But what can be done in Rainmeter is more than enough for now and it looks spectacular:

Now I'm looking at how to show only that screen in fullscreen, maybe there is an application that can zoom in on a previously selected region.

During the weekend I will continue exploring possibilities.

Astro.-

 

Wow .. Great Display !

Very nice :)

But there are problems with the image overlay .. the current BIG problem is  that the overlay function does not understand how to work with transparencies !

The red tint works because it has no transparency.

What needs to happen is the incoming image needs to be converted to 4 channel with transparency and merged with the overlay which has transparency..  

I'll look into how to do this ..